Plans and limits
Two plans. Free runs on your machine and every capability is included: one project, every browser, every store, manual operation. Pro is $29 per workspace per month and buys unattended operation: builds from webhooks, releases, and the API while nobody watches. Capability is never gated, only automation.
Attended means started from the console. Webhook, release, and api triggers are unattended.
When a limit is reached the platform refuses; it never bills past the plan. The invoice is $0 or $29 per workspace per month, nothing else.
The limits
Every enforced limit:
| Limit | Free | Pro |
|---|---|---|
| Projects | 1 | 10 |
| Build dispatches per month | 100 | 350, fair use |
| Build dispatch burst, per 60-minute sliding window | 12 | 30 |
| AI generations per day | 50 | 200 |
| AI generations per month | 300 | 2000 |
Share link lifetime is not in that table because it is not a plan limit. A shared preview lives 30 days on both plans, and a share of a build the project already published carries a token of at most 168 hours on both. See Share an unpublished build.
The free monthly dispatch number is being finalized against real usage and may change.
When it fails
When a cap fires, the refusal names its reason: monthly_cap, burst_cap, kill_switch, counter_unavailable, or unattributed. See promote gates for the full list.
Operators hold kill switches that can pause build dispatch and AI generation platform wide. When one is engaged, the refusal says so; nothing on your side is wrong.
Next
- Usage and limits for what is metered and where the counters live
- Promote gates for every refusal
- Access tokens for the tokens that attribute
apidispatches